Docstoc

Figure 1

Document Sample
Figure 1 Powered By Docstoc
					        Urban Climate Studies:
 applications for weather, air quality,
          and climate change
               Prof. Robert Bornstein
               Dept. of Meteorology
             San Jose State University
                 San Jose, CA USA
              pblmodel@hotmail.com

                   Presented at
                Tel Aviv University
                    8 Jan 2009

Funding sources: USAID-MERC, TCEQ, NASA, NSF, SCU
                                                    1
               OVERVIEW
• URBAN CLIMATE
  – WHY STUDY IT
  – ITS CAUSES
  – ITS IMPACTS
• CALIFORNIA COASTAL COOLING
  – DATA
  – ANALYSIS
• URBAN ATMOSPHERIC MODELS
  – FORMULATION
  – APPLICATIONS (HOUSTON, ATLANTA, ISRAEL)
• FUTURE EFFORTS

                                              2
      URBAN WEATHER ELEMENTS:
    battles between conflicting effects
•   VISIBILTITY: decreased
•   TURBULENCE: increased (mechanical & thermal)
•   PBL NIGHT STABILITY: neutral
•   FRONTS (synoptic & sea breeze): slowed
•   TEMP: increased (UHI) or decreased
•   PRECIP: increased (UHI) or decreased
•   WIND SPEED (V): increased or decreased
•   WIND DIRECTION: con- or divergence
•   THUNDERSTORMS: triggered or split
                                                   3
   HUMAN-HEALTH IMPACTS OF
       URBAN CLIMATE
• > UHI  THERMAL STRESS
• > PRECIP ENHANCEMENT  FLOODS
• > URBAN INDUCED INVERSIONS 
     POLLUTED LAYERS
• > TRANSPORT & DIFF PATTERNS FOR
  – POLLUTION EPISODES
  – EMERGENCY RESPONSE (i.e., TOXIC
    RELEASES)

                                      4
 NEW URBAN CLIMATE: CAUSES
• GRASS & SOIL 
    CONCRETE & BUILDINGS 
    ALTERED SURFACE HEAT FLUXES
• FOSSIL FUEL CONSUMPTION 
    ATMOSPHERIC POLLUTION AND HEAT
• ATM POLLUTION
    ALTERED SOLAR & IR ENERGY


                                     5
           St. Louis nocturnal PBL:
   warm near-neutral, polluted urban-plume
      vs. rural stable surface-inversion

                                 0F




  Tmin

Tmax
              urban-plume

  inversion




                            Clark & McElroy (1970):   6
    Urban effects on wind speed
• FAST LARGE-SCALE (i.e., SYNOPTIC) SPEEDS 
       SMALL UHI 
       URBAN SFC ROUGHNESS (Z0) INDUCED
       DECELERATION
•   SLOW SYNOPTIC SPEEDS
       LARGE UHI 
       INWARD-DIRECTED ACCELERATION
•   CRITICAL SPEED ~ 3-4 m/s (FOR NYC & London)



                                                  7
NYC DAYTIME ∆V (z)




       urban   rural




                       8
URBAN EFFECTS ON WIND DIR
• FAST SYNOPTIC SPEED WEAK UHI 
    URBAN BUILDING-BARRIER EFFECT 
    FLOW DIVERGES AROUND CITY
• SLOW SYNOPTIC SPEED  LARGE UHI 
    LOW-p  CONVERGENCE INTO CITY
• MODERATE SYNOPTIC SPEED 
    CONVERGENCE-ZONE ADVECTED TO
    DOWNWIND URBAN-EDGE
                                      9
  NOCTURNAL UHI-INDUCED SFC-CONFLUENCE:
            otherwise-calm synoptic flow 
confluence-center over urban center of Frankfurt, Germany




                                                     10
  Weak cold-frontal (N to S) passage over NYC
a. Hourly positions (left)
b. At 0800 EST (right): T, q, & SO2 z-profile-changes
   showed lowest 250 m of atm not-replaced, as front
   “jumped” over city




          See 



                                                   11
  URBAN IMPACTS ON PRECIP
• INITATION BY THERMODYNAMICS (at SJSU)
  – LIFTING FROM
     • UHI CONVERGENCE
     • THERMAL & MECHANICAL CONVECTION
  – DIVERGENCE FROM BUILDING BARRIER EFFECT
• AEROSOL MICROPHYSICS
  – SLOWER SECONDARY DOWNWIND ROLE
  – METROMEX & PROF. D. ROSENFELD (HUJI)


                                           12
NYC two-summer daytime-average thunderstorm-precip radar-echoes
(σ’s from uniform-distribution) for cases: all, convective, & moving




                                                    Formed over city


                                     splitting
                                     case



                     Split by city
                                                                 13
             Dispersion effects
• Vertical diffusion limited by urban-induced
        elevated inversions (next slide)
•   Transport: 3-D effects of urban-induced flow-
        modifications
•   Convergence-zones effects due to
     – Urban effects
     – Sea breezes


                                                    14
 Urban-induced nocturnal elevated inversion-I traps home-heating emissions
 Power plant plume is trapped b/t urban-induced inversions I & II
 Inversion III is regional inversion  poor estimate of mixing depth




                                                                   Plume

                                            Home-heating
                                              Sources



                                                                              15
        California Coastal-Cooling
        (to appear, J. of Climate, 2009)

• Global & CA observations generally show
  – asymmetric warming (more warming for Tmin than
    for Tmax) (next graph)
  – acceleration since mid-1970s
• CA downscaled global-modeling (next map)
  – done onto 10 km grids
  – shows summer warming that decreases towards
    the coast (but does not show coastal cooling)
                                                16
Not much change from mid-
40s to mid-70s, when values
started to again rapidly rise




                            17
   Statistically down-scaled (Prof. Maurer, SCU) 1950-2000
Summer (JJA) IPCC temp-changes (0C) show warming rates that
   decrease towards coast; red dots are COOP sites used in
           present study & boxes are study sub-areas




                                                        18
            Current Hypothesis
INCREASED GHG-INDUCED
INLAND TEMPS
INCREASED (COAST TO
INLAND) PRESSURE & TEMP
GRADIENTS
INCREASED SEA BREEZE FREQ,
INTENSITY, PENETRATION,
&/OR DURATION 
COASTAL AREAS SHOULD
SHOW COOLING SUMMER
DAYTIME MAX TEMPS (i.e., A
REVERSE REACTION)

NOTE:
NOT A TOTALLY ORIGINAL
IDEA                            19
  Results 1: SoCAB 1970-2005 summer (JJA) Tmax warming/
cooling trends (0C/decade); solid, crossed, & open circles show
   stat p-values < 0.01, 0.05, & not significant, respectively

       ?                        ?




                                                          ?




                                                              20
             Results 2: SFBA & CV 1970-2005 JJA
Tmax   warming/cooling trends (0C/decade), as in previous figure


                                                   ?

               ?




                                          ?




                                                            21
Results 3: JJA Temp trends; all CA-sites
                                  • LOWER TRENDS
                                    FROM 1950- 70
                                    (EXCEPT FOR TMAX)
                                  • Curve b: TMIN HAD
                                    FASTEST RISE (AS
                                    EXPECTED)
                                  • Curve c: TMAX HAD
                                    SLOWEST RISE; IT IS
                                    A SMALL-∆ B/T BIG
                                    POS VALUE & BIG
                                    NEG-VALUE (AS IN
                                    ABOVE 2 GRAPHS)
                                  • CURVE a: TAVE THUS
                                    ROSE AT MID RATE
                                  • Curve d: DTR (diurnal
                                    temp range) THUS
                                    DECREASED (AS TMAX
                                    FALLS & TMIN RISES)
                                                   22
Significance of these all-CA Trends
• HIGHER TRENDS FROM 1970-2005 
    FOCUS NEEDED ON THIS PERIOD
•   TMIN HAS FASTER RISE 
    ASSYMETRIC WARMING IN LITERATURE
•   BUT TMAX HAS SLOWER RISE, BECAUSE IT IS A
    SMALL DIFFERENCE B/T BIG POS-VALUE & BIG
    NEG-VALUE (AS SEEN IN ABOVE SPATIAL PLOTS)
•   TAVE & DTR ARE ALSO THUS “CONTAMINATED”
•   NEXT 2 SLIDES THUS SHOW SEPARATE TRENDS
    FOR CA COASTAL AND INLAND AREAS

                                            23
    Result 4: JJA Tave, Tmin, Tmax, & DTR TRENDS FOR
     INLAND-WARMING SITES OF SoCAB & SFBA
a
                                      Curve b: TMIN
                                      INCREASED
                                      (EXPECTED)
b                                     Curve c: TMAX HAD
                                      FAST RISE;
                                      (UNEXPECTED),
                                      COULD BE DUE TO
c                                     INCREASED UHIs OR
                                      INCREASED DOWN-
                                      SLOPE FLOWS

                                      CURVE a: TAVE THUS
d                                     ROSE AT MID RATE

                                      Curve d: DTR THUS
                                      INCREASED (AS TMAX
                                      ROSE FASTER THAN
                                      TMIN ROSE         24
    Result 5: JJA Tave, Tmin, Tmax, & DTR TRENDS FOR
     COASTAL-COOLING SITES OF SoCAB & SFBA
a
                                    Curve b: TMIN ROSE
                                    (EXPECTED)

                                    Curve c: TMAX HAD COOL-
b                                   ING (UNEXPECTED MAJOR
                                    RESULT OF STUDY)

                                    CURVE a: TAVE THUS
                                    SHOWED ALMOST NO
c                                   CHANGE, AS FOUND IN
                                    LIT.), AS RISING Tmin &
                                    FALLING Tmax CHANGES
                                    ALMOST CANCELLED OUT

d                                   Curve d: DTR THUS DE-
                                    CREASED, AS TMIN ROSE &
                                    TMAX FELL


                                                         25
Note IPCC 2001 does show cooling over
        Central California!!




                                        26
Significance of above Coastal-Cooling
     and Inland-Warming trends
• CA ASSYMETRIC WARMING IN LITERATURE IS
    HEREIN SHOWN TO BE DUE TO COOLING TMAX
    IN COASTAL AREAS & CONCURRENT WARMING
    TMAX IN INLAND AREAS
•   PREVIOUS CA STUDIES
    – DID NOT LOOK SPECIFICALLY AT SUMMER DAYTIME
      COASTAL VS. INLAND VALUES HAVE
    – THEY THUS REPORTED CONTAMINATED TMAX, TAVE, &
      DTR VALUES
    – THEY, HOWEVER, ARE NOT INCONSISTENT WITH
      CURRENT RESULTS, THEY ARE JUST NOT AS
      DETAILED IN THEIR ANALYSES & RESULTS
                                                 27
Result 6. JJA 1970-2005 2 m Tmax trends for 4 pairs of
   urban (red, solid) & rural (blue, dashed) sites
Notes:
1. All sites are near the
   cooling-warming border
2. UHI-TREND (K/DECADE)
   = absolute sum b/t
   warming-urban &
   cooling-rural trends
  a. SFBA sites
    > Stockton
       (0.38 + 0.17 = 0.55)
    > Sac. (0.49)
 b. SoCAB sites
    > Pasadena (0.26)
    > S. Ana (0.12)
                                                   28
          Notes on JJA daytime
            UHI-trend results

• Faster growing cities (not shown) had
  faster growing UHIs
• As no coastal sites showed warming Tmax
  values, calculations could only be done at
  these four pairs (at the inland boundary
  b/t the warming and cooling areas)
• Coastal sites would have cooled even
  more w/o their (assumed) growing UHIs

                                               29
        BENEFICIAL IMPLICATIONS OF
            COASTAL COOLING
• NAPA WINE AREAS MAY NOT GO EXTINCT
    (REALLY GOOD NEWS!) (next map)
•   ENERGY FOR COOLING MAY NOT INCREASE AS
    RAPIDLY AS POPULATION (next graph)
•   LOWER HUMAN HEAT-STRESS RATES
•   OZONE CONCENTRATIONS MIGHT CONTINUE
    TO DECREASE, AS LOWER MAX-TEMPS MEAN
    REDUCED
    – ANTHROPOGENIC EMISSIONS
    – BIOGENIC EMISSIONS
    – PHOTOLYSIS RATES
                                         30
NAPA WINE AREAS MAY NOT GO EXTINCT DUE TO ALLEGED
   RISING TMAX VALUES, AS PREDICTED IN NAS STUDY




                                               31
Result 7: Peak-Summer Per-capita Electricity-Trends


                                 Down-trend at cooling
                                Coastal: LA (blue) & Pasadena
                                (pink, 8.5%/decade)

                                > Up-trend at warming
                                inland Riverside (green)



                                 Up-trend at warming Sac
                                 & Santa Clara

                                 Need:
                                  detailed energy-use
                                 data for more sites
                                  to consider changed
                                 energy efficiency
                                                           32
Future Coastal-Cooling Efforts (PART 1 OF 2)
• EXPAND (TO ALL OF CA & ISRAEL?)
  – ANALYSIS OF OBS (IN-SITU & GIS)
  – URBANIZED MESO-MET (MM5, RAMS, WRF) MODELING
• SEPARATE INFLUENCES OF CHANGING:
  – LAND-USE PATTERNS RE
     • AGRICULTURAL IRRIGATION
     • URBANIZATION & UHI-MAGNITUDE
  – SEA BREEZE:
     INTENSITY, FREQ, DURATION, &/OR PENETRATION
• DETERMINE POSSIBLE “SATURATION” OF SEA-
 BREEZE EFFECTS FROM
     • FLOW-VELOCITY & COLD-AIR TRANSPORT
     • AND/OR STRATUS-CLOUD EFFECTS ON LONG- & SHORT-WAVE
      RADIATION
                                                      33
 POSSIBLE FUTURE EFFORTS (PART 2 OF 2)
• DETERMINE CUMULATIVE FREQ DISTRIBUTIONS
 OF TMAX VALUES, AS
  – EVEN IF AVERAGE TMAX DECREASES,
  – EXTREME VALUES TMAX MAY STILL INCREASE (IN
    INTENSITY &/OR FREQUENCY)
• DETERMINE CHANGES IN LARGE-SCALE ATM
 FLOWS:
  – HOW DO GLOBAL CLIMATE-CHANGE EFFECT POSITION
    & STRENGTH OF: PACIFIC HIGH & THERMAL LOW?
  – THESE TYPES OF CLIMATE-CHANGES ARE THE
    ULTIMATE CAUSES OF TEMP AND PRECIP CHANGES


                                                 34
       OUR GROUP‟S MESO-MODELING EXPERIENCE
• SJSU (MM5 & uMM5)
   –   Lozej (1999) MS: SFBA winter wave cyclone
   –   Craig (2002) MS: Atlanta UHI-initiated thunderstorm (NASA)
   –   Lebassi (2005) MS: Monterey sea breeze (LBNL)
   –   Ghidey (2005) MS: SFBA/CV CCOS episode (LBNL)
   –   Boucouvula (2006a,b) Ph.D.: SCOS96 episode (CARB)
   –   Balmori (2006) MS: Tx2000 Houston UHI (TECQ)
   –   Weinroth (2009) PostDoc: NYC-ER UDS urban-barrier effects (DHS)
• SCU (uRAMS)
   – Lebassi (2005): Sacramento UHI (SCU)
   – Lebassi (2009) Ph.D.: SFBA & SoCAB coastal-cooling (SCU)
   – Comarazamy (2009) Ph.D.: San Juan climate-change & UHI (NASA)
• Altostratus (uMM5 & CAMx)
   –   SoCAB (1996, 2008): UHI & ozone (CEC)
   –   Houston (2008): UHI & ozone (TECQ)
   –   Central CA (2008): UHI & ozone (CEC)
   –   Portland (current): UHI & ozone (NSF)
   –   Sacramento (current): UHI & ozone (SMAQMD)


                                                                         35
SJSU IDEAS ON GOOD MESO-MET
          MODELING
MUST CORRECTLY REPRODUCE:
– UPPER-LEVEL Synoptic/GC FORCING FIRST:
   pressure (“the” GC/Synoptic driver) 
   Synoptic/GC winds
– TOPOGRAPHY NEXT:
   min horiz grid-spacing 
   flow-channeling
– MESO SFC-CONDITIONS LAST:
   temp (“the” meso-driver) & roughness 
   meso-winds
                                            36
Case 1: ATLANTA UHI-INITIATED STORM: OBS GOES &
    PRECIP (UPPER) & MM5 w’s & precip (LOWER)




                                              37
    Recent Meso-met Model Urbanizations
• Need to urbanize momentum, thermo, & TKE
     – Surface & SfcBL Diagnostic-Eqs.
     – PBL Prognostic-Eqs. (not done in NCAR uWRF)
•   Start: veg-canopy model (Yamada 1982)
•   Veg-param replaced with GIS/RS urban-param/data
     – Brown and Williams (1998)
     – Masson (2000)
     – Martilli et al. (2001) in TVM/URBMET
     – Dupont, Ching, et al. (2003) in EPA/MM5
     – Taha et al. („05, „08a,b,c) [& Balmori et al. („06)]: his uMM5
       uses improved urban dynamics, physics, parameterizations,
       & inputs


                                                                 38
 From EPA uMM5:        Within Gayno-Seaman
Mason + Martilli (by    PBL/TKE scheme
    Dupont)
                                         New




                                             39
 But, uMM5 needs extra GIS-derived inputs
             as f (x, y, z, t)
 land-use (38 categories)
 roughness heights z0 (see next slide)
 anthropogenic heat
 building heights
 paved-surface 2-D fractions
 building H to W, wall-plan, & impervious-area
     2-D ratios
 building frontal, plan, & rooftop 3-D area densities

                                                  40
S. Stetson: Houston GIS/RS zo inputs



                         But, values are too
                         large, as they were
                         f(h) & not f(ơh)

                         h = building height




                                             3
                               Values up to 41 m
uMM5 for Houston: Balmori (2006)
Goal: Accurate urban/rural temps & winds for
 Aug 2000 O3 episode via
  – uMM5
  – Houston LU/LC & urban morphology
    parameters
  – TexAQS2000 field-study data
  – USFS urban-reforestation scenarios 
    UHI & O3 changes



                                           42
uMM5 Simulation period: 22-26 August 2000
• Model configuration
     – 5 domains: 108, 36, 12, 4, 1 km
     – (x, y) grid points:
        (43x53, 55x55, 100x100, 136x151, 133x141
     – full-s levels: 29 in D 1-4 & 49 in D-5; lowest ½ s-level=7 m
     – 2-way feedback in D 1-4
•   Parameterizations/physics options
       > Grell cumulus (D 1-2)       > ETA or MRF PBL (D 1-4)
       > Gayno-Seaman PBL (D-5) > Simple ice moisture,
       > urbanization module NOAH LSM > RRTM radiative cooling
•   Inputs
       > NNRP Reanalysis fields, ADP obs data
       > Burian morphology from LIDAR building-data in D-5
       > LU/LC modifications (from Byun)

                                                              43
1-km grid, uMM5 Houston UHI: 8 PM, 21 Aug




      UHI


            Bay           UHI



                  Gulf




MM5 UHI (2.0 K)
                            uMM5 UHI (3.5 K)
                                           44
UHI-Induced Convergence: obs vs. uMM5




       C

                           C




Krieged Obs             uMM5 output

                                        45
           Base-case (current)
 min
            veg-cover (0.1’s)
            urban min (red)
            rural max (green)



           Modeled changes of
  max
            veg-cover (0.01’s)
increase    Urban-reforestation
               (green)
           Rural-deforestation
              (purple)
                             46
Run 12 (urban-max reforestation) minus Run 10 (base case) 
               near-sfc ∆T at 4 PM shows that:
            reforested central urban-area cools &
          surrounding deforested rural-areas warm




                           warmer


                  cooler




              warmer

                                                        47
  DUHI(t): Base-case minus Runs 15-18




           RURAL
  URBAN
                                       Max-impact of –0.9 K on
                                        a 3.5 K noon-UHI, of which
                                       1.5 K was from uMM5


• UHI = Urban-Box minus Rural-Box
• Runs 15-18: Urban re-forestation scenarios
• DUHI = Run-17 UHI minus Run-13 UHI 
      max effect (green line)
• Reduced UHI  lower max-O3 (not shown) 
      EPA emission-reduction credits  $ saved             48
RAMS, MM5, & CAMx SIMULATIONS OF
 MIDDLE-EAST O3 TRANSBOUNDARY
           TRANSPORT

      E. Weinroth1,2, S. Kasakseh1,3
         M. Luria2, R. Bornstein1
            1San Jose State Univ.
        2Hebrew Univ. Jerusalem, Israel
 3Applied Research Institute Jerusalem (ARIJ),

            Bethlehem, West Bank

           In Atmos. Environ. (2008)




                                                 49
            USAID-MERC project (2000-)
• Involves scientists from Palestinian Territories, Israel,
    USA (& now Jordan and Lebanon)
•   Objectives accomplished:
    – Installation of environmental stations in West Bank & Gaza
      (and now Jordan & Beirut)
    – Preparation of environmental databases (SJSU web page)
    – Field campaigns during periods of poor air quality (Prof. Luria)
    – Application of numerical models for planning
        • RAMS & MM5 (Kasakseh 2007) meso-met
        • CAMx photochemical air-quality (Weinroth et al. 2007 in
          Atmos. Environ.)

                                                                   50
                 Night obs of sfc flow:
                 3-AM LST (00 UTC)
        H
            Flow Dir: weak down-slope off
              coastal-mountains at
             Coastal plain: offshore (to W)
               from W-facing slopes
             Haifa Pen. (square): offshore
            (to E ) from E-facing slopes
    L        Inland sites: directed inland (to
               E) from E-facing slopes

L           Low-O3
            generally <40 ppb)
            Haifa still at 51 ppb

                                          51
             Day Obs: 1200 NOON LST
    L
            Winds:
            Reversed
            Stronger: up 6 m s-1

             Coastal plain: Onshore/upwind,
               from SW
        H
             Inland sites: Channeling (from
              W) in corridor (box; focus of
              modeling) from Tel-Aviv to
              J. area (at Modiin site).
L       H
            Higher daytime O3
             max at Mappil, 66 ppb
             2nd max at Modiin, 63 ppb

                                      52
                             MM5 Configuration

 Version 3.7
 3 domains
     – 15, 5, 1.67 km Grid Spacings
     – 59 x 61, 55 x 76, 58 x 85 Grid
         Points
   32 σ-levels
     – up to 100 mb
     – first full σ-level at 19 m
   Lambert-conformal map projection
    (suitable for mid lat regions)
   Two-way nesting
   5-layer soil-model
   Gayno-Seaman PBL
   Simulations
     – End: 00 UTC, 3 Aug
     – Start: 00 UTC, 29 July
   Single CPU , LINUX



                                                 53
MM5 Domain-3 winds (m/s) at 1100 LST on 1 Aug „97
red lines = topo heights (m); yellow line = sea breeze
 front; note reverse upslope-flow & channeling to J.


               Sea


                                            Max



                                       J.




                                 Max

                                                   54
   Same, but at 2300 LST; where yellow line = land
  breeze front; note down-slope flow; still inland
directed flow in inland areas & still channeling to J.


               Sea



                                            Max


                                       J.




                                 Max


                                                     55
    Mid-east Obs vs. MM5: 2 m temp (Kasakech ‟06 AMS)

                        First 2 days show GC/Syn trend not in MM5,
                        as MM5-runs had no analysis nudging
                                            Obs



                                                 Run 1




                                                                  Run 4:
              July 29                August 1               AugustReduced
                                                                   2
                                                obs               Seep-soil
                                                               MM5:Run 4 T
             July 31                   Aug 1                      Aug2




Standard-MM5 summer night-time min-T,
                                                                           56
But lower input deep-soil temp  better 2-m T results  better winds  better O3
Obs vs. MM5: wind speed (m/s)
                                         Run 3
                        OBS




 July 31     August 1         August 2




                                            57
 RAMS/CAMx (left) O3 vs. airborne obs (right) at 300 m:
 > Secondary-max: over J. in obs; due to coastal N-S highway
 > Primary-max: in Jordan (no obs); due to Hadera


                                              Airborne obs
             Jerusalem            Irbid,
                                  Jordan
             Hadera
O3 ppb


     0
             Power      .
    0-20

   20-40      Plant
   40-60

   60-70

   70-80

   80-90

   90-95

   95-105

   105-120      1 Aug, 1500 LST

                                                         58
     Overall Modeling Lessons
• > Models can‟t be
  – assumed to be perfect (i.e., model user vs. modeler)
  – used as black boxes
• > Need good large-scale forcing model-fields
• > If obs are not available, OK to make reasonable
     educated estimates, e.g., for rural
  – deep-soil temp
  – soil moisture
• > Need data to compare with simulated-fields
• > Need good urban
  – morphological data
  – urbanization schemes
                                                           59
Thanks for listening!

    Questions?




                        60

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:1
posted:11/10/2011
language:English
pages:60